Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Apr 5;3(4):100293.
doi: 10.1016/j.xgen.2023.100293. eCollection 2023 Apr 12.

Escape from oncogene-induced senescence is controlled by POU2F2 and memorized by chromatin scars

Ricardo Iván Martínez-Zamudio  1   2 , Alketa Stefa  1   3 , José Américo Nabuco Leva Ferreira Freitas  4   5   6 , Themistoklis Vasilopoulos  1   3 , Mark Simpson  1 , Gregory Doré  7 , Pierre-François Roux  8 , Mark A Galan  9 , Ravi J Chokshi  10 , Oliver Bischof  6 , Utz Herbig  1
Affiliations

Escape from oncogene-induced senescence is controlled by POU2F2 and memorized by chromatin scars

Ricardo Iván Martínez-Zamudio et al. Cell Genom. .

Abstract

Although oncogene-induced senescence (OIS) is a potent tumor-suppressor mechanism, recent studies revealed that cells could escape from OIS with features of transformed cells. However, the mechanisms that promote OIS escape remain unclear, and evidence of post-senescent cells in human cancers is missing. Here, we unravel the regulatory mechanisms underlying OIS escape using dynamic multidimensional profiling. We demonstrate a critical role for AP1 and POU2F2 transcription factors in escape from OIS and identify senescence-associated chromatin scars (SACSs) as an epigenetic memory of OIS detectable during colorectal cancer progression. POU2F2 levels are already elevated in precancerous lesions and as cells escape from OIS, and its expression and binding activity to cis-regulatory elements are associated with decreased patient survival. Our results support a model in which POU2F2 exploits a precoded enhancer landscape necessary for senescence escape and reveal POU2F2 and SACS gene signatures as valuable biomarkers with diagnostic and prognostic potential.

Keywords: AP-1; OIS; OIS escape; Oct-2; POU2F2; SACS; cellular senescence; colorectal cancer; oncogene-induced senescence; senescence-associated chromatin scars.

PubMed Disclaimer

Conflict of interest statement

The authors declare no competing interests.

Figures

None
Graphical abstract
Figure 1
Figure 1
Transcriptional landscape of OIS escape (A) Diagram describing the time-resolved multiomic profiling approach used to define the gene-regulatory mechanisms of escape from OIS. (B) Averaged self-organizing maps (SOMs) of transcriptomes of three independent experiments of GM21 fibroblasts undergoing escape from OIS (horizontal) and time-matched empty vector control (EV) (vertical) expressed as logarithmic fold change. The transcriptional landscape of each stage of the OIS escape process is highlighted. P, proliferation; S, senescence; T, transition; E, escape. (C and D) PCA projection plots showing individual (C) and averaged (D) transcriptional trajectories of DEGs of three biologically independent OIS escape time-series experiments. (E) Heatmap showing the color-coded modules (n ≥ 300 genes per module) of DEGs of GM21 fibroblasts in P, S, T, and E. Average of three biologically independent experiments is shown. (F) Functional overrepresentation analysis map showing significant associations of the MSigDB hallmark gene sets for each module described in (E). Circle fill is color coded according to the false discovery rate (FDR)-corrected p value from a hypergeometric distribution test. Circle size is proportional to the percentage of genes in each MSigDB gene set. (G) Volcano plot showing DEGs in cells after OIS escape, relative to EV proliferating cells. (H) Functional overrepresentation analysis map showing significant associations of the MSigDB hallmark gene sets with genes up and down in post-senescent cells relative to proliferating control cells. Up, 118 genes; down, 150 genes. (I) Heatmap showing the expression levels of SASP genes from the OIS-specific turquoise module (118 genes predicted to be secreted) of cells in P, S, T, and E. Examples are shown in the insets. Data are averaged from three independent experiments.
Figure 2
Figure 2
Enhancer remodeling dictates transcriptional transitions and OIS escape (A) Arc-plot visualization of the chromatin state transitions at indicated stages as cells enter and escape from OIS. The width of the edge is proportional to the number of 200 bp bins undergoing a given chromatin transition. (B) Histogram visualization of 200 bp bins undergoing the top 20 most frequent chromatin state transitions during the escape from OIS. (C) Asymmetric biplot of the correspondence analysis (CA) between representative chromatin state transitions and gene expression modules. Nine representative and best-projected (squared cosine > 0.5) chromatin state transitions are shown. Red lines arising from the origin indicate the projections of gene expression modules (from Figure 1E). Statistical significance was calculated using a chi-squared test. (D) Distance distribution of each peak undergoing chromatin state transitions to the TSS of DEGs in modules described in Figure 1E. Bin size, 20 kb. (E) Integration of eight select chromatin state transitions (top pictograms, days after H-RASG12V expression) with nearby gene expression output (row Z score boxplots). Centerline of boxplots indicates the median. Edges correspond to the first and third quartiles. Whiskers extend from the edges to 1.5× the interquartile range to the highest and lowest values. Chromatin state transition data (A–D) was computed from two independent immunoprecipitations per time point per histone modification from pooled chromatin from 10 biologically independent experiments. Gene expression data (D) are the average of three biologically independent time series.
Figure 3
Figure 3
Organized waves of TF activity define OIS escape (A) Heatmap showing differential TF chromatin binding activity (row Z score) at enhancers at each stage of OIS escape. Only expressed TFs were considered in the analysis. Annotations on left show the number of bound instances per TF and their gene expression (TXN) category (i.e., constitutively expressed [black] or differentially regulated according to the module color code shown in Figure 1E). Insets: chromatin binding activity of representative TF families. (B and C) TF co-binding matrices at open (B) and closed (C) enhancers in cells entering and escaping from OIS. All binding instances across time points were collapsed onto the matrix and clustered using Ward’s aggregation criterion. Corresponding q values were projected onto the clustering and are color coded based on significance calculated using a hypergeometric distribution test. TF footprinting (A–C) and differential chromatin binding activity (A) were performed on pooled ATAC-seq datasets from two biologically independent time-series experiments.
Figure 4
Figure 4
Hierarchical TF networks define transcriptional dynamics of OIS escape (A–D) Effector TF networks for each gene expression module (Figure 1E). TFs (nodes) are represented as circles. Oriented edges (arrows) connecting nodes indicate that at least 15% of the regions bound by a given TF in the bottom and core layers were bound by the interacting TF in the core and top layers, respectively, at the same or previous time points. Strongly connected components (SCCs) are represented as a single node to facilitate visualization. The fill color of the node’s inner circle is based on the normalized dynamicity of TFs. The fill color of the outer ring indicates whether the TF is constitutively expressed (black) or belongs to a transcriptomic module (Figure 1E). The node’s size is proportional to the bound regions by a given TF. Each network has three layers: (1) the top layer with no incoming edges, (2) the core layer with incoming and outgoing edges, and (3) the bottom layer with no outgoing edges. Networks were generated from pooled ATAC-seq datasets from two biologically independent time series.
Figure 5
Figure 5
POU2F2 is necessary for OIS escape (A) Immunoblot analysis of indicated proteins in GM21 fibroblasts undergoing OIS escape. γ-tubulin and Ponceau are loading controls. One representative blot of three biologically independent experiments is shown. (B) Genome-wide POU2F2 enhancer-binding during OIS escape and time-matched controls. Footprinting was performed on pooled ATAC-seq datasets from two biologically independent experiments. (C) Expression heatmap of POU2F2 gene targets within gene expression modules identified in Figure 1E. (D) Functional overrepresentation analysis map showing significant associations of the MSigDB hallmark gene sets with each POU2F2 gene target in the indicated modules. (E) Western blot analysis of POU2F2 expression in RAS-expressing GM21 fibroblasts expressing non-targeting and POU2F2-targeting shRNAs. γ-tubulin and Ponceau are loading controls. One representative blot of four biologically independent experiments is shown. (F) Cumulative population doubling plot at the OIS (S, days 8–16), transition (T, days 18–25), and escape (E, days 26–35) phases of GM21 fibroblasts expressing non-targeting and POU2F2-targeting shRNAs. Population doublings within each time interval per stage of four biologically independent experiments were included in the analysis. ∗p = 0.0412 using a two-tailed unpaired Student’s t test. (G) Quantitation of 5-Ethynyl-2′-deoxyuridine (EdU)-positive GM21 fibroblasts expressing non-targeting and POU2F2-targeting shRNAs at senescence (S) and OIS escape (E) stages. Data are from three biologically independent experiments. ∗∗p = 0.0088, ∗p = 0.0173, using a two-tailed unpaired Student’s t test. (H) Representative immunofluorescence micrographs of EdU incorporation at the OIS and escape phases in GM21 fibroblasts expressing non-targeting and POU2F2-targeting shRNAs. (I) GSEA showing normalized enrichment score (NES) plots and FDR values for POU2F2 predicted target genes (from C) in transcriptomes of pLKO- and shPOU2F2-expressing cells at day 14 after H-RASG12V expression. Statistical evaluation of GSEA results was based on a non-parametric Kolmogorov-Smirnov test. (J) Intersections of positively and negatively regulated POU2F2 target genes within the predicted gene set defined in (C). (K and L) Expression heatmaps of positively and negatively POU2F2-regulated genes. (M) Functional overrepresentation analysis map showing significant associations of the MSigDB hallmark gene sets with positively (+ve) and negatively (−ve) regulated POU2F2 target genes. Only genes concordantly modulated by both POU2F2-specific shRNAs were considered in the analyses (I–M). (N) Histogram plot showing the normalized CDKN2A reads in cells as described in (I). ∗Adjusted p = 3.05 × 10−5, Wald test, relative to empty-vector cells in the E phase. The average of two biologically independent experiments is shown. (O) Immunoblot analysis of indicated proteins in H-RASG12V-expressing GM21 fibroblasts expressing non-targeting and POU2F2-targeting shRNAs. γ-tubulin and Ponceau were used as loading controls. Numbers on top of the CDKN2A/p16INK4a blot show a densitometric fold increase relative to pLKO at each time point. One representative blot of three biologically independent experiments is shown.
Figure 6
Figure 6
POU2F2 drives an inflammatory gene expression program in colorectal cancer (A) Kaplan-Meier plot showing overall survival comparing colorectal cancer (CRC) patients with low (blue) and high (red) normalized POU2F2 mRNA expression in primary tumors from the TCGA-COADREAD dataset as defined in (B). Statistical significance was conducted using a log-rank test. The p value and n for each POU2F2 expression group are indicated. Vertical red and blue lines show each group’s 50% survival probability. (B) Normalized POU2F2 expression in the TCGA-COADREAD tumor samples. Red hashed line indicates expression cutoff used for survival analysis (0.72 Fragments Per Kilobase of transcript per Million mapped reads upper quartile (FKPM-UQ). (C) Genome-wide meta profiles of POU2F2 binding at accessible chromatin in normal colon tissue (NC) and CRC. Footprinting was performed on 12 NC (five donors) and 38 tumor DNase-seq and ATAC-seq datasets. (D) GSEA map showing significant associations of KEGG gene sets with a POU2F2 escape gene signature in CRC samples. Circle fill is color coded according to the adjusted p value based on a non-parametric Kolmogorov-Smirnov test. Circle size is proportional to the number of POU2F2 target genes within each KEGG gene set (rows). POU2F2 CRC gene targets were identified using gene expression data from 101 NC and 631 primary tumors deposited in the TCGA using the Xena Differential Gene Expression Analysis Pipeline. POU2F2 targets in CRC were then overlapped with POU2F2 targets in GM21 fibroblasts that escaped from OIS, and the union of these two sets was analyzed. (E and F) Representative micrographs at 10× (top; scale bar, 50 μm) and 40× (bottom; scale bar, 10 μm) original magnifications of hematoxylin and immunohistochemical analysis of POU2F2 expression in adjacent non-tumor biopsy and adenocarcinoma. (G) GSEA showing normalized enrichment score (NES) plots and FDR values for the POU2F2 escape gene signature in a comprehensive analysis of CRC transcriptomes (GEO: GSE14333, GSE33113, GSE37892, GSE39852) profiled after surgery and classified into relapse (n = 262) and no-relapse (n = 720) groups. Statistical evaluation of GSEA results was based on a non-parametric Kolmogorov-Smirnov test.
Figure 7
Figure 7
SACSs define post-senescent cells and are present in colorectal cancer (A) Heatmap showing the chromatin accessibility modules at enhancers per time point as cells enter and escape from OIS. The analysis identified 27,117 differentially accessible regions. Biologically independent replicates are numbered. (B) Chow-Ruskey diagram showing intersections and disjunctive unions of differentially accessible chromatin regions at each time point measured as cells enter and escape from OIS relative to the empty-vector control (day 0). The black center circle depicts 1,926 senescence scars. (C) True/false emission heatmap visualization of differentially accessible genomic regions at each chromatin accessibility module at enhancers for each time-point pairwise comparison, as indicated at the bottom of the heatmap. SACSs span horizontally for all time-point comparisons relative to day 0 (five leftmost lanes of the heatmap). (D) Histogram visualization of 200 bp windows within senescence scars undergoing the top 10 most frequent chromatin state transitions. (E and F) Functional overrepresentation analysis maps showing significant associations of the MSigDB hallmark (E) and KEGG (F) gene sets with genes associated with SACSs in each chromatin accessibility module. Circle fill is color-coded according to the FDR-corrected p value from a hypergeometric distribution test. Circle size is proportional to the ratio of genes in the database set (MSigDB or KEGG) found within the genes associated with SACSs in each chromatin accessibility module. The numbers of biologically independent experiments per time point are indicated in (A). (G–I) Rank plots showing the summed binding instances of TFs in accessible chromatin, SACSs in CRC, and SACSs in post-senescent cells. (J) Normalized meta profiles of SACSs in individual normal colon and CRC samples. (K) Heatmap showing the modules (a–i) of differentially expressed genes in normal colon (n = 24), adenomas (n = 44), and adenocarcinomas (n = 34) from GEO: GSE20916. (L) Functional overrepresentation analysis map showing significant associations of POU2F2 and SACS gene signatures in the modules described in (K). (M) GSEA showing NES plots and FDR values for SACS gene signature in a comprehensive analysis of transcriptomes of samples of patients with CRC (GEO: GSE14333, GSE33113, GSE37892, GSE39852) profiled after surgery and classified into relapse (n = 262) and no-relapse (n = 720) groups. SACS-associated genes in CRC were identified using gene expression data from 101 normal colon and 631 primary tumors deposited in TCGA using the Xena Differential Gene Expression Analysis Pipeline.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Serrano M., Lin A.W., McCurrach M.E., Beach D., Lowe S.W. Oncogenic ras provokes premature cell senescence associated with accumulation of p53 and p16INK4a. Cell. 1997;88:593–602. doi: 10.1016/s0092-8674(00)81902-9. - DOI - PubMed
    1. Collado M., Serrano M. Senescence in tumours: evidence from mice and humans. Nat. Rev. Cancer. 2010;10:51–57. doi: 10.1038/nrc2772. - DOI - PMC - PubMed
    1. Saab R. Senescence and pre-malignancy: how do tumors progress? Semin. Cancer Biol. 2011;21:385–391. doi: 10.1016/j.semcancer.2011.09.013. - DOI - PubMed
    1. Gray-Schopfer V.C., Cheong S.C., Chong H., Chow J., Moss T., Abdel-Malek Z.A., Marais R., Wynford-Thomas D., Bennett D.C. Cellular senescence in naevi and immortalisation in melanoma: a role for p16? Br. J. Cancer. 2006;95:496–505. doi: 10.1038/sj.bjc.6603283. - DOI - PMC - PubMed
    1. Fujita K., Mondal A.M., Horikawa I., Nguyen G.H., Kumamoto K., Sohn J.J., Bowman E.D., Mathe E.A., Schetter A.J., Pine S.R., et al. p53 isoforms Delta133p53 and p53beta are endogenous regulators of replicative cellular senescence. Nat. Cell Biol. 2009;11:1135–1142. doi: 10.1038/ncb1928. - DOI - PMC - PubMed